Joanna Yu, PhD

FROM THE NATIONAL NETWORK OF DEPRESSION CENTERS

One of the biggest challenges in treating depression is the ability to select the most appropriate and effective treatment for a particular individual from among the many available options. This is a significant public health issue, since the first treatment selected is effective only about 

Although there are excellent clinical guidelines to address the management of a major depressive episode, the evidence is based on average rates of response in group data and not on individual characteristics or objective biological measures.

While the use of biomarkers in psychiatry holds great potential, large multimodal studies with standardized methods of data collection, longitudinal assessments, and collaboration among groups are necessary for discovery and subsequent validation of findings. Researchers have adopted this approach in an effort to stratify populations of depressed individuals based on their response to various treatments including pharmacotherapy, cognitive behavioral therapy (CBT), and neurostimulation.

Furthermore, standardized data collection using common measures within and across studies is required to help tease apart the heterogeneity of depression and thus identify treatment response subtypes. These measures can range from cognitive tests to blood-based protein assays as well as functional and structural MRI brain scans. Once these rich multidimensional datasets are integrated, the hope is that big data analytics will uncover clinically meaningful biological differences that can be translated into biomarkers.

Canadian Biomarker Integration Network in Depression

 (CAN-BIND) is one of several multisite initiatives generating rich integrated datasets designed to identify biosignatures to inform treatment selection.

 CAN-BIND’s standardized multimodality platform approach to data collection began with a 16-week open label standardized antidepressant treatment study called CAN-BIND-1. This internationally recognized research and education program has since expanded to include more than 10 clinical studies spanning 8 clinical research sites across Canada.

CAN-BIND researchers and study participants are working together to generate a large harmonized dataset that will span diverse interventions and specific populations, including adolescents at risk for developing mental health disorders and those who have attempted suicide. Collection of high quality data across CAN-BIND sites and studies is enabled by the Ontario Brain Institute’s (OBI) large-scale web-based neuroinformatics platform known as the Brain-Centre for Ontario Data Exploration (

OBI, a provincially funded, not-for-profit research organization, created Brain-CODE to support collection, centralized storage, federation, sharing, and analysis of 

 (eg, structured interview and self-report measures, MRI scans, and molecular assays) across several brain disorders.

 In partnership with the Indoc Consortium,

 OBI created Brain-CODE to breakdown silos, promote team science, address commonalities across diseases and disorders, and harness expertise across diverse disciplines; their combined goal is to improve the lives of patients diagnosed with neurological and mental health disorders. In this article, we highlight how CAN-BIND and its major sponsor, OBI, are working together through Brain-CODE to drive forward the discovery of depression biomarkers (

CAN-BIND is one of 5 OBI-sponsored Integrated Discovery Programs mandated to take a truly transdisciplinary research approach to advancing brain health.

 To maximize discovery from multimodal data collection, harmonized data points and established quality control processes are necessary to ensure sustained collection of high quality data and data integration. To facilitate this, CAN-BIND established platforms led by experts in each of the domains of interest: clinical outcomes, neuroimaging scans, electroencephalography (EEG) recordings, molecular assays involving standardized biospecimen collection, and mobile-health (m-health) technologies.

Within this piece, we briefly describe standardization challenges, solutions provided by Brain-CODE, and potential effects within and across platforms. The hope is that these rich multidisciplinary datasets will contribute to the development and validation of tools and technology to help with self-management, real-time monitoring, and early detection and prevention of depression and relapses.

 While depression severity scales, such as the Montgomery-Asberg Depression Rating Scale (MADRS) and Hamilton Depression Rating Scale (HAM-D), are commonly used by research clinicians to assess patients, more nuanced behavioral and self-report tools to assess such variables as sleep and alertness, cognitive deficits, and anhedonia provide additional clinical information that could inform treatment selection. As such, 10 clinician-administered and 19 self-report questionnaires, all of which are based on theoretical and clinical utility, are part of the standardized battery of clinical assessments in the CAN-BIND-1 study to help identify predictors of response to standard antidepressant treatment regimen.

Data collection is enabled by electronic data capture tools, including Research Electronic Data Capture (REDCap)

 on Brain-CODE. This minimizes the need for paper forms and potential error related to data entry. The data gathered from this large battery of questionnaires can help identify key clinical characteristics that predict treatment outcomes and allow evidence-based decisions to effectively guide treatment selection and monitor progress. In addition to reducing the number of measures moving forward, these results also can be used to inform the development of user-friendly mobile health applications that will enable remote collection and real-time monitoring capabilities.

Multisite neuroimaging standardized data collection.

 In comparison to collecting clinical outcomes, collecting and standardizing brain scans is more costly and more technically difficult, especially when accounting for different MRI machines across sites. While this poses significant challenges for data integration, the additional biological information gained holds great promise for new discoveries and is representative of real world clinical practice settings. CAN-BIND has taken on this challenge and implemented multiple quality control steps, including regular phantom scans that serve as a common reference to facilitate downstream analysis from different machines. In addition, semi-automated quality control procedures delivered through Brain-CODE’s imaging informatics software platform called Stroke Patient Research Recovery Database (SPReD),

 which is a part of the Extensible Neuroimaging Analysis Toolkit (XNAT), provides additional regular checks to ensure collection of high quality brain imaging data across CAN-BIND sites and studies.

CAN-BIND is collecting structural and functional brain scans, both resting state and during various tasks, in an attempt to identify different depression subgroups based on brain connectivity and structure.

 These neuroimaging findings will contribute to a better understanding of the structural and/or functional changes associated with depression and treatment response, which in turn may inform treatment selection and facilitate a more targeted delivery. Hopefully this also will lead to the development of more cost-effective and time-efficient proxy tests to shorten the path to recovery.

 Brain activity measured by EEG may serve as a proxy biomarker for neuroimaging. With recent technological advances, portable EEG devices make it possible for assessments to be completed quickly and in community settings. Further research and development is needed to define measurements and create algorithms to predict and/or monitor treatment response.

As a first step, CAN-BIND has developed standardized methodology to collect and analyze EEG data from dedicated research machines across sites and studies.

 This has yielded encouraging results as pilot EEG data from the CAN-BIND-1 study at baseline and 2 weeks post-antidepressant treatment have shown significant utility toward predicting treatment response.

 To investigate the possibility of enhancing sensitivity and specificity of predicting treatment outcome, follow-up analyses involving integration with MRI, clinical, and molecular data are being conducted.

Molecular and mobile health technology data.

 has the capacity to handle collection and storage of a wide array of molecular data ranging from inflammatory makers to functional genomic and genetic data. LabKey serves as a file repository and allows for secure transfer and storage of CAN-BIND’s m-health data collected from wearables and mobile applications.

While individual platforms within CAN-BIND will contribute to expanding our knowledge of depression, the real strength and goal of the network is to carry out integrative data analyses across data collection platforms. To help prepare data for complex large-scale analytics, harmonized data are aggregated and interactive visual dashboards are created to query metadata and generate data packages on Brain-CODE.

Having completed platform specific data quality control and assurance, cleaning, and curation processes for the CAN-BIND-1 study, CAN-BIND is currently conducting integrative analyses across data platforms. These analyses involve collaborations to increase analytics capabilities in machine learning, network biology, computational psychiatry, and other data science methodologies. This further highlights the multidisciplinary team effort and vast array of expertise required to advance depression biomarker discovery. Brain-CODE is able to support these collaborative analyses by providing personalized virtual group workspaces with readily available access to such analytic packages as R Studio and MATLAB.

For the CAN-BIND-1 study, CAN-BIND’s major goal is to identify biomarkers that predict treatment response to standard pharmacotherapy. A longer term goal is to expand this precision-treatment approach to other interventions. To this end, CAN-BIND is generating a large harmonized dataset across treatments like CBT and neurostimulation therapies such as repetitive transcranial magnetic stimulation (rTMS).

In addition to integration across CAN-BIND studies, there are opportunities to collaborate and federate data with other international depression initiatives, such as the Establishing Moderators and Biosignatures of Antidepressants Response for Clinical Care for Depression (EMBARC) initiative. This should expedite discovery and validation of findings. Hopefully, this collaborative team effort will lead to readily accessible biological-based tests that can accurately and efficiently guide treatment selection for depression.

CAN-BIND, through Brain-CODE, will also have the opportunity to form linkages with other databases to both expand and augment datasets for further discovery. For example, Brain-CODE has developed a customized subject registry tool that collects an encrypted version of study participants’ Ontario Health Insurance Plan number; this enables linkage with the Institute for Clinical Evaluative Sciences health administrative database. In turn, this will allow CAN-BIND to link research and health care utilization data to conduct health economic analyses, including evaluation of the effects of new biomarkers and interventions and, ultimately, generate evidence to influence policy.

CAN-BIND, through its unique partnership with OBI and Brain-CODE, is also poised to advance our understanding of depression in other neurological disorders. For instance, in addition to CAN-BIND, OBI has supported the development of research programs in epilepsy, cerebral palsy, neurodevelopmental disorders (eg, autism), and neurodegenerative diseases (eg, Alzheimer disease, Parkinson disease).

To facilitate collaboration among the five existing OBI research programs, several common data elements (CDEs) were established using a Delphi consensus process with participation from all research programs. These CDEs reflect existing international standards established by such organizations as the National Institute of Neurological Disorders and Stroke and Clinical Data Interchange Standards Consortium. They include standardized instruments to collect demographic, depression, sleep, anxiety, quality of life, and other clinical questionnaire-based data. This paves the road for analysis of comorbid symptoms and common pathways related to depression and mental health (

To maximize outcomes, Brain-CODE is designed with data sharing, collaborative promotion, and open science in mind. Much thought and careful planning has gone into ensuring appropriate Brain-CODE governance, privacy, and security measures are in place to allow data from Brain-CODE to be shared, not only within and across OBI’s research programs, but also eventually with third parties to support data mining, discovery, and innovation.

Big data holds great potential for multidisciplinary innovations toward precision treatment and improving brain health. From CAN-BIND, this includes the use of portable brain activity monitoring devices, mobile applications, and blood based molecular assays as biomarker readouts, likely in combination with novel analytic algorithms, to improve management and treatment of depression.

This article is based on Drs Brendan Behan’s and Joanna Yu’s presentation at the 2017 Annual National Network of Depression Center (NNDC) Conference.

NNDC’s goal is to foster collaborative research to advance scientific discovery and drive forward improved care for patients with depression and mood disorders. NNDC is an American-based initiative and is a content partner of 

is Neuroinformatics Manager, Centre for Depression and Suicide Studies, St. Michael’s Hospital; 

 is Program Lead, Informatics and Analytics, Ontario Brain Institute; 

 is Manager, Informatics and Analytics, Ontario Brain Institute and Chief Operating Officer and Director of Clinical Research, Indoc Research; 

is Vice President, Research and Informatics, Ontario Brain Institute; 

 is Professor of Psychiatry, University of Michigan, Ann Arbor; 

is Program Manager, Centre for Depression and Suicide Studies, St. Michael’s Hospital and Assistant Professor of Psychiatry, University of Toronto; 

 is Director, Centre for Depression and Suicide Studies, St. Michael’s and Professor of Psychiatry, University of Toronto.

Rush AJ. Star-D: Lessons learned and future implications. 

 Vaccarino AL, Dharsee M, Strother SC, et al. Brain-CODE: a secure neuroinformatics platform for management, federation, sharing and analysis of multi-dimensional neuroscience data. 

Stuss D. The Ontario Brain Institute: completing the circle. 

Lam RW, Milev R, Rotzinger S, et al. Discovering biomarkers for antidepressant response: protocol from the Canadian biomarker integration network in depression (CAN-BIND) and clinical characteristics of the first patient cohort. 

 Fonseka TM, MacQueen GM, Kennedy SH. Neuroimaging biomarkers as predictors of treatment outcome in major depressive disorder. 

 Faranak F, Atluri S, Frehlich M, et al. Standardization of electroencephalography for multi-site, mutli-platform and multi-investigator studies: insights from the Canadian Biomarker Integration Network in Depression. 

 Baskaran A, Farzan F, Milev R, et al. The comparative effectiveness of electroencephalographic indices in predicting response to escitalopram therapy in depression: a pilot study. 

Articles

One of the biggest challenges in treating depression is the ability to select the best treatment for a particular individual from among the many available options.

Author | Joanna Yu, PhD

Articles

Big Data for Depression